Using the predicted protein sequence of AtPS1, two ESTs were identified on DFCI Potato Gene Index (http://compbio.dfci.harvard.edu/tgi) named TC194262 and DN590600 corresponding to FHA and PINc domains, respectively. The primers designed on the above mentioned ESTs allowed to isolate in diploid potato a 3 Kbp genomic clone (PSL1) lacking UTR regions. In order to complete genomic sequence of PSL1, by querying Potato Genome Sequencing Consortium Database (http://www.potatogenome.net/index.php) and SOL Genomic Network (http://solgenomics.net) we retrieved a 3 Kbp region of S. phureja v3 scaffold PGSC0003DMS000001829 (phuPSL) and a tomato BAC clone AC211085.1 (SlPSL1) sharing 97% and 93% of sequence identity with PSL1, respectively. Primers designed on the tomato sequence were used to isolate a potato 5.3 Kbp PSL1 genomic region spanning 2 Kbp upstream of the start codon to about 200 bp downstream of the stop codon [GenBank:HQ418834]. In silico gene prediction showed a structure of PSL1 composed by six exons and five introns (Figure 1A) and a 2.4 Kbp hypothetical PSL1 cDNA (PSL1_pred) with a GC content of 42% encoding a 92.5 kDa protein of 823 aminoacids (pPSL1_pred).

Given the meiotic function of AtPS1, potato PSL1 cDNA was isolated from pre-bolting buds. Seven different cDNAs ranging from 2.3 to 2.7 Kbp were isolated including PSL1a of 2.4 Kbp corresponding to PSL1_pred. The sequence alignment between cDNAs and the genomic PSL1 indicated the presence of different groups of related PSL sequences encoded by more than one locus (Figure 1A). In order to investigate the relationship between the different PSL cDNAs, we conducted a phylogenetic analysis that suggested the existence of three different loci named PSL1, PSL2 and PSL3 (Figure 1B). On the basis of sequence similarity, we assigned PSL1a, PSL1b and PSL1c cDNAs [GenBank:HQ418835, GenBank:HQ418836 and GenBank:HQ418837] to genomic PSL1, PSL3a, PSL3b and PSL3c cDNAs [GenBank:HQ418839, GenBank:HQ418840 and GenBank:HQ418841] to PSL3 and the last cDNA to PSL2 [GenBank:HQ418838]. Moreover, the distance measured as the number of different nucleotides among the seven cDNAs showed that PSL2 was more similar to PSL3 than to PSL1 (Figure 1C). The nucleotide comparison between the predicted phuPSL cDNA and PSL1a, PSL2 and PSL3c cDNAs showing a similarity of 98.5%, 99.1% and 99.6%, respectively, suggested that phuPSL locus corresponds to PSL3 being the differences explained by the different genetic background.

Based on the sequences of cloned PSL cDNAs, PSL1b and PSL1c, PSL3a and PSL3b are alternative splicing forms of PSL1 and PSL3 since they retained complete or partial introns causing the formation of premature stop codons (PTCs) (Figure 1A). Moreover, the cloned PSL2 cDNA showed a PTC caused by the retention of the second intron. As a consequence, all the predicted PSL proteins, except pPSL1a and pPSL3c, had truncated or lacking PINc domain (Figure 1D).

In order to evaluate whether the alternative splicing PSL variants were possible target of degradation through NMD we calculated the distance between PTCs and the successive exon-exon junction. Being this distance more than 50-55 nt according to Nagi and Maquat 14 we could consider PSL1b, PSL1c, PSL2, PSL3a and PSL3b as target of NMD.

In order to investigate the evolution of PSL family, a phylogenetic analysis was performed by a search on Interpro (http://www.ebi.ac.uk/interpro) for proteins with both FHA and PINc domains. Interestingly, these domains were contemporary present only in plants, except for the multidrug-efflux transporter NAEGRDRAFT_78193 from the amoeboid Naegleria gruberi.

PSL1a sequence was blasted against the Phytozome v6 database (http://www.phytozome.net) that contains the genomic sequences of 22 organisms and against the SOL Genomics Network containing the tomato genome assembly. Afterwards, 25 sequences of predicted PSL proteins were collected from 19 different organisms (Additional File 1), since Physcomitrella patens, Ricinus communis, Volvox carteri, Zea mays and Chlamydomonas rehinardtii seemingly lack PSL genes. The sequences were then aligned and the Maximum-Likelihood phylogenetic tree is shown in Figure 2. The distribution of PSLs is in agreement with the known phylogenetic relationships between species among dicots and monocots. Moreover, one PSL locus was found in the analyzed plant species, except for Glycine max. Indeed, four different PSL loci were identified in soybean and three of them encode alternative transcripts. Differently from potato, the alternative transcripts of soybean retain PINc domain being the splicing sites located at the 3'-end of mRNA (Figure 3).

In order to assess the conservation degree of PSLs, we predicted and compared the secondary structure of FHA and PINc domains using the SMART (http://smart.embl-heidelberg.de) and PSIPRED (http://bioinf.cs.ucl.ac.uk/psipred) tools. It is reported that FHA domain is 80-100 aminoacid (aa) long folded into a 11-stranded beta sandwich, which sometimes contains small helical insertions between the loops connecting the strands. However, in silico analysis of FHA displays only 8 beta-strands (b-strands) out of 11 including the residues involved in phosphopeptide recognition and stabilisation of domain architecture 15 . Using the above mentioned tools on yeast RAD53p [NCBI:6325104], a well characterized FHA containing protein 12 , the identified FHA region was 52 aa covering 6 beta-strands (data not shown). As shown in Figure 4, the length of the predicted FHA region in our dataset was 53 aa except for Brassica rapa (51 aa) and for Glycine max PSL2 (68 aa). While the majority of FHA domains showed 6 beta-strands, Brassica rapa FHA was predicted to be composed of 4 consecutive beta-strands followed by an alpha helical region. The group of monocots, Brachypodium distachyon, Oryza sativa and Sorghum bicolor, as well as dicot Glycine max (GmPSL1) showed a helical insertion between the 2nd and the 3rd beta-strand. In other species this helical insertion was predicted but at a low confidence value as estimated by PSIPRED. Afterward, we compared the active sites in yRAD53p with those present in PSL^FHA domains through the protein alignment showed in Figure 5. It can be observed that glycine-5, arginine-6, serine-21 and histidine-24 in FHA domain are perfectly conserved. The arginine-19 seems to be absent in all plant sequences. In the analyzed species, asparagine-60 showed a substitution with histidine, characterized by a different polarity, except for soybean PSL4 glutamine-60 and Brassica rapa that seems to lack this site. Asparagine-66 is mostly substituted with the similar polar serine. Instead, AlPSL1, AtPSL1, GmPSL3 and GmPSL4 exhibited arginine-66 with different polarity while BrPSL1 glutammine-66 and MtPSL1 cysteine-66 both showing similar polarity of asparagine.

The analysis of PINc domain in our dataset started with the prediction of its secondary structure in human SMG6 (hSMG6) [UniProt:Q86US8]. It is reported that hSMG6^PINc (hSMG6) is 182 aa folded into a 5-stranded parallel beta-sheets that is highly twisted and alpha-helices arranged on both sides of each beta-sheet for a total of 6. Three aspartate residues are essential for PINc activity in hSMG6, while a threonine or a serine in the sequence (T/S)XD might be involved in catalytic role on the basis of similarity with other PINc domains 16 17 . SMART predicted a PINc domain of 152 aa lacking the first and the last alpha-helices while PSIPRED predicted 4 beta-sheets and 4 alpha helices in hSMG6 (data not shown). In our dataset the PINc domain ranged from a minimum of 123 aa in Cucumis sativus to a maximum of 162 aa in Brachypodium distachyon (Figure 6). In our prediction, the number of beta-sheets ranged from 3 in GmPSL3 and CsPSL1 to 9 in CpPSL1. The number of predicted alpha-helices ranged from 3 in CsPSL1, OsPSL1 and PSL3c to 7 in BdPSL1 and CpPSL1. Afterward, we compared the active residues of PINc in PSL proteins. The alignment between PINc domain of PSL proteins and hSMG6 is shown in Figure 7 (refer to Additional file 2: Alignment of PSL^PINc and hSMG6^PINc residues to look at raw alignment). BrPSL1, SmPSL1 and PSL3c show the three expected aspartate residues at positions 6, 194 and 233 of the alignment. SiPSL1 and SbPSL1 have a substitution of two aspartate residues with asparagine-6, glutammate-194. When a substitution occurs, the aspartate-194 is mostly replaced with glutammate-194 except for CpPSL1 showing a different polarity residue (lysine-194) and MgPSL1 showing an alyphatic residue (alanine-194). The aspartate-233 is widely conserved except in BdPSL1, OsPSL1 and SiPSL1 where it is substituted with a serine-233. Most of the PSL proteins show a catalytic serine-231 in the described pattern SXD (where X can be D, N, E or S in this study) instead of threonine-231 in hSMG6, PtPSL1, SiPSL1, SmPSL1 and VvPSL1. GmPSL3 and SbPSL1 are the only proteins lacking the C-terminal extremity of PINc interrupting at leucine-203 and lysine-227 respectively thus missing the catalytic threonine/serine-231 and the aspartate-233.

In order to test for presence of positive selection at individual amino acid codons, the site specific models implemented in DataMonkey 2010 webserver (http://www.datamonkey.org) 18 were used. The Integrative Selection Analysis of FEL, SLAC and FER alghoritms evidenced no significant positive selected sites (dN-dS > 0). Conversely, 217 significant negative selected sites (dN-dS < 0) were identified. The region between FHA to PINc domains included few negative selected sites which were mostly located near the functional domains. The codons encoding the highly conserved active sites glycine-5, arginine-6, serine-21 and histidine-24 in FHA and aspartate-6, serine/threonine-231 and aspartate-233 in PINc were also subjected to negative selection (Figure 8). These observations support the results obtained through sequence alignment and evidence an occurrence of negative pressure upon non-synonymous mutations in PSL genes. In particular, the regions next to the functional domains and the codons for active sites were affected.

A previously described 30 diploid clone of potato (named T710) coming from hybridization between Solanum tuberosum haploid USW3304 (2n = 2x = 24) and S. chacoense (2n = 2x = 24) has been used to isolate PSL genes.

Plant genomic DNA was isolated from leaves using the DNeasy Plant Mini Kit (QIAGEN http://www1.qiagen.com). Bacterial plasmid DNA was isolated using QIAprep Spin Miniprep Kit (QIAGEN). Total RNA from prebolting buds was extracted using the RNeasy Plant Mini Kit (QIAGEN) and then treated with DNase I (Invitrogen, http://www.invitrogen.com) to remove residual genomic DNA. Primer pairs for cloning designed with PRIMER3 (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3.cgi) are listed in Additional file 3: Primers used in this study.

PSL1 genomic fragment was amplified by PCR with primers PS_F1 and PS_R1 and full sequence with primers PS_F2 and PS_R2. The coding region of PSL transcripts was amplified by RT-PCR by using SuperScriptTM III Reverse Transcriptase (Invitrogen) with an oligo- dT_12-18 (Invitrogen). The cDNA was subjected to PCR with PS_CF and PS_CR primers (Additional File 3). PCR products have been cloned into PCR2.1 with T/A Cloning kit (Invitrogen). Nucleotide sequencing was carried out by Eurofins MWG Operon sequencing service (Germany).

PSL1 gene structure and cDNA predictions were carried out using FGENESH online tool (http://linux1.softberry.com/berry.phtml?topic=fgenesh&group=programs&subgroup=gfind) selecting Tomato as organism. In silico translation of cDNA sequences was carried out using the Expasy Translation tool (http://www.expasy.ch/tools/dna.html).

The PSL protein sequences were identified and collected by TBLASTN 31 search against the Phytozome v6 database (http://www.phytozome.net/), SGN (SOL Genomic Network, http://sgn.cornell.edu/) and Potato Genome Sequencing Consortium (PGSC, http://potatogenomics.plantbiology.msu.edu/index.php?p=blast).

The PSL protein sequences were firstly aligned by MUSCLE 32 using the default settings of MEGA5 33 . The best model for Maximum Likelihood phylogeny analysis was chosen testing all the available models in ProtTest version 2.4 34 with slow optimization strategy and selecting the one with highest AICc value. The evolutionary history was then inferred by using the Maximum Likelihood method according to Jones et al. w/freq. model 35 . The bootstrap consensus tree was inferred from 100 replicates 36 . Initial tree(s) for the heuristic search were obtained automatically as following. When the number of common sites is <100 or less than one fourth of the total number of sites, the maximum parsimony method was used, otherwise BIONJ method with MCL distance matrix was used. A discrete Gamma distribution was used to model evolutionary rate differences among sites (4 categories, +G, parameter = 1.3900). The rate variation model allowed for some sites to be evolutionarily invariable ([+I], 9.0566% sites). All ambiguous positions were removed for each sequence pair. Evolutionary analyses were conducted in MEGA5.

The locations of FHA and PINc domains within PSL genes were detected using SMART 37 . Secondary structure prediction was performed using domains from PSL sequences as input into the PSIPRED secondary structure prediction server 38 . The program MUSCLE 32 was used to do multiple sequence alignments of FHA and PINc domains in PSL proteins, yRAD53p [NCBI:6325104] and hSMG6 [UniProt:Q86US8]. Conservation of phosphothreonine-binding residues in FHA and of RNAse activity residues in PINc were determined by alignment with yRAD53p and hSMG6, respectively.

The coding sequences of PSLs were obtained from the databases reported in Additional File 1. The terminal codon was manually removed, then the codon alignment was performed by MUSCLE 32 using the default settings of MEGA5 33 (Tamura et al., personal communication). To select the best substitution model we used JmodelTest 0.1.1 39 40 with default settings. The GTR 41 model was selected as the best fitting as evidenced by AICc values.

The codon alignment was then uploaded on DataMonkey 2010 server 18 42 and dN-dS evaluated using GTR as substitution model, and SLAC (default settings except for Global dN/dS value = estimated with CI), FEL and REL 43 alghoritms. Codons subjected to evolutive pressure were identified with Integrative Selection Analysis selecting significance levels for SLAC p < 0.1, FEL p < 0.1 and REL Bayes Factor < 50. Only codons with a significant dN-dS value according to all the three methods were reported.